Parquet Diff
Two versions of a dataset, one question: what actually changed? Join both files on a key column and see the added, removed and changed rows — plus any schema drift — before you ship the new version.
Nothing is uploaded: the comparison runs locally via WebAssembly, so it is safe for confidential data.
How it works
Byte-level diff tools report every Parquet file pair as different — row-group layout, compression and encoding all change the bytes without changing the data. This tool compares the data itself: both files are joined on the key you pick (inside an in-browser DuckDB database), rows are classified as added, removed, changed or unchanged, and changed rows show exactly which cells differ. Schema changes — added, removed or re-typed columns — are detected instantly from the file metadata, before any rows are read.
When you need this
Re-ran a pipeline and want to check the output still matches yesterday's? Reviewing a data hand-off and need to confirm what a colleague changed? Migrating jobs and validating that the new code writes the same rows? The usual answer is a throwaway pandas script — load both files, merge on a key, compare columns. This page replaces that script with a drag & drop, and because everything runs client-side, it works with files you are not allowed to upload anywhere.